Update FINAL_TRAINING_REPORT.md

FINAL_TRAINING_REPORT.md

@@ -19,10 +19,10 @@ training:
   batch_size: 64
   
   # DataLoader optimizations
-  num_workers: 4  # Number of parallel data loading processes (0 for main thread only)
-  pin_memory: true  # Pin memory for faster CPU->GPU transfer
-  persistent_workers: true  # Keep workers alive between epochs (only if num_workers > 0)
-  prefetch_factor: 2  # Number of batches to prefetch per worker (only if num_workers > 0)
+  num_workers: 4
+  pin_memory: true
+  persistent_workers: true
+  prefetch_factor: 2
   
   # Smooth curriculum over digit lengths (optional)
   curriculum:
@@ -73,8 +73,8 @@ evaluation:
   final_samples_per_combination: 1000  # Samples per digit combination for final evaluation
 
 logging:
-  interval_examples: 5000  # Log training loss every N examples (was: 250 batches * 32 batch_size = 8000 examples)
-  save_path: "models/minimal_example_lora"  # Model will be uploaded to HuggingFace as your-username/minimal-addition-lora
+  interval_examples: 5000  # Log training loss every N examples
+  save_path: "models/minimal_example_lora"
 ```
 
 ## Training Summary